Search papers, labs, and topics across Lattice.
This paper investigates agentic misalignment in large language models (LLMs) by employing chain-of-thought (CoT) monitoring to identify harmful actions arising from goal conflicts. The authors introduce INTENT-AS-A-TOOL, which integrates intent-targeted tools to provide a nuanced signal of the model's commitment to specific behaviors, enhancing the granularity of intent tracking. The results indicate that this method not only complements existing CoT monitoring but also facilitates real-time intervention by revealing critical decision-making steps during reasoning processes.
Intent-targeted tools reveal that LLMs often signal harmful intentions before executing risky actions, enabling real-time intervention strategies.
As large language models (LLMs) are deployed as autonomous agents, safety failures increasingly involve consequential actions. We study agentic misalignment, where agents take harmful actions under goal conflicts and pressures. Using chain-of-thought (CoT) monitoring, we find that harmful execution is often preceded by intent signals in reasoning. However, post-hoc CoT labels are too coarse to show how intent changes during generation. We introduce INTENT-AS-A-TOOL, an approach that adds intent-targeted tools to give the model a dedicated channel for expressing commitment to a target behavior. The probability of calling an intent tool provides a judge-free, fine-grained signal of the model's tendency to pursue that behavior. Our results show that INTENT-AS-A-TOOL complements CoT monitoring, expands post-hoc CoT labels into dense trajectories, and identifies critical steps for online intervention. These findings suggest that action preferences are useful for tracking agentic misalignment during reasoning. Our code and data are accessible: https://github.com/RebeccaZhang22/intent-as-a-tool.
